CN114328950A - Power distribution network fault disposal knowledge map construction and intelligent aid decision making system and method - Google Patents
Power distribution network fault disposal knowledge map construction and intelligent aid decision making system and method Download PDFInfo
- Publication number
- CN114328950A CN114328950A CN202111465857.7A CN202111465857A CN114328950A CN 114328950 A CN114328950 A CN 114328950A CN 202111465857 A CN202111465857 A CN 202111465857A CN 114328950 A CN114328950 A CN 114328950A
- Authority
- CN
- China
- Prior art keywords
- fault
- fault handling
- distribution network
- entity
- plan
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to the technical field of power systems, and particularly relates to a power distribution network fault handling knowledge map construction and intelligent aid decision making system and method; the technical scheme is as follows: the power distribution network fault treatment system comprises a fault treatment data preprocessing module, a power distribution network fault treatment knowledge extraction module, a fault treatment knowledge map construction module and a fault treatment decision reasoning module, wherein the fault treatment data preprocessing module is connected with the power distribution network fault treatment knowledge extraction module, the power distribution network fault treatment knowledge extraction module is connected with the fault treatment knowledge map construction module, and the fault treatment knowledge map construction module is connected with the fault treatment decision reasoning module. And realizing intelligent extraction of fault disposal knowledge, establishing a plan event intention identification model through TextCNN, judging the plan event execution intention, and finishing automatic execution of a fault disposal plan according to the identified plan event operation components and the plan event intention, namely finishing intelligent disposal of the distribution network fault.
Description
Technical Field
The invention belongs to the technical field of power systems, and particularly relates to a power distribution network fault handling knowledge graph construction and intelligent aid decision making system and method.
Background
The alternating current and direct current hybrid large power grid in China is formed, single faults are easy to cause cascading faults when being not timely disposed, and higher requirements are put forward on the fault disposal effect and the normalization. The power distribution network is a final link closely connected with the user terminal, is a key point for ensuring the power supply reliability, and has no pressure for a dispatcher to deal with faults in the face of new challenges of data diversity, complexity, islanding, high degree of combination with task scenes and the like. The existing regulation and control system can comprehensively acquire and monitor the running state of the distribution network, and in the case that the distribution network fault does not form a whole set of service flow processing system, the decision is made by the experience of a dispatcher, and the pre-programmed rules and plans lack intelligent electronic and modeling means, so that the auxiliary decision can not be effectively provided for the dispatcher in time when the fault occurs, and a large amount of idle operation is caused. In conclusion, it is needed to improve the cognitive intelligence level of the data perception of the power distribution network and to standardize the fault handling capacity by learning the texts such as the current regulation and control rules and the plans to form a knowledge model.
Disclosure of Invention
The invention overcomes the defects in the prior art and provides a power distribution network fault handling knowledge graph construction and intelligent aid decision making system and method.
In order to solve the technical problems, the invention adopts the technical scheme that:
the power distribution network fault handling knowledge map construction and intelligent aid decision making system comprises a fault handling data preprocessing module, a power distribution network fault handling knowledge extraction module, a fault handling knowledge map construction module and a fault handling decision reasoning module, wherein the fault handling data preprocessing module is connected with the power distribution network fault handling knowledge map extraction module, the power distribution network fault handling knowledge map construction module is connected with the fault handling knowledge map construction module, and the fault handling knowledge map construction module is connected with the fault handling decision reasoning module.
The power distribution network fault disposal knowledge graph construction and intelligent assistant decision method comprises the following steps:
s1, the fault handling data preprocessing module cleans the structured data and the unstructured data based on the regular expression;
s2, extracting fault disposal knowledge in a fault disposal plan by a power distribution network fault disposal knowledge extraction module based on a hybrid neural network to obtain fault disposal key points and key point relations, and extracting BIO marks and events of the power distribution network fault disposal plan by using verbs as trigger words;
s3, the fault handling knowledge map construction module stores the extracted fault handling plan handling knowledge into a Neo4j database according to information categories such as plan titles, operation modes, control measures, fault reasons, control requirements, after-fault influences and the like, modularizes the whole fault handling knowledge map, and judges the operation intention of the combined plan event through a text convolution neural network;
s4, the fault handling decision reasoning module takes a fault handling plan knowledge graph as a handling engine, and intelligent handling of faults is completed under the action of the reasoning engine by sensing the real-time running state of the power distribution network.
Preferably, in step S1, the structured data is table information from a relational database, and includes plant stations, lines, transformers, buses, switches, breakers, power company address books, and new energy plant stations, and the unstructured data is text information, and includes scheduling rules, fault handling plans, and fault handling cases.
Preferably, the step S2 includes the steps of:
s21, classifying each event component in the fault handling plan event text according to natural grammar, namely entity categories, and dividing the fault handling plan entity into 6 categories of execution subject, verb trigger, execution object, execution supplement, execution subject and execution condition, 1 non-entity label and 7 labels in total;
s22, adopting ALBERT-CRF to identify entity types in the fault handling plan, combining 6 fault handling plan entities of subject execution, verb trigger words, object execution, shape execution and condition execution to form a complete event according to grammatical relations, and finishing the extraction of the fault handling plan event;
s23, converting the fault handling plan entity into a calculable word vector by using an ALBERT dynamic word vector model, realizing coding of the fault handling plan entity, sending the coded word vector into entity information label types of predicted characters of an ALBERT network parameter adjusting layer, and finally outputting a global optimal label sequence by taking sentences as units on a CRF layer.
Preferably, the initial part of the execution subject entity is represented as 'B-sub', and the rest part of the execution subject entity is represented as 'I-sub'; the initial part of the verb trigger entity is represented as 'B-act', and the rest of the verb trigger entity is represented as 'I-act'; the beginning part of the execution object entity is represented as 'B-obj', and the rest part of the execution object entity is represented as 'I-obj'; the initial part of the execution object is represented as 'B-obj _ m', and the rest part of the execution object is represented as 'I-obj _ m'; the beginning part of the executive shape language entity is represented as 'B-adv', and the rest part of the executive shape language entity is represented as 'I-adv'; the beginning part of the execution conditional entity is denoted as "B-con", the rest of the execution conditional entity is denoted as "I-con", and the non-entity component is denoted as "O".
Preferably, the ALBERT is a lightweight BERT model, the parameter dimension of a BERT hidden layer is reduced by decomposing word vectors, and the decomposed word vectors can be hidden to the hidden layer according to requirements.
Preferably, the function of the CRF layer is to predict the optimal label scores of each category from a global perspective, and the calculation formula is as follows:
in the formula:
are the adjacent state transition matrix distribution values in the CRF,
is the fraction value of the LSTM model output vector, and after normalizing the fraction value, the probability formula of the model is obtained:
in the formula: YX represents all sequence tags;
the CRF optimizes the label sequence by solving the logarithm of the probability, and finally obtains the sequence with the highest score as the final predicted output sequence through the maximum likelihood estimation, wherein the calculation formula is as follows:
in the formula:
represents the score of the entire predicted sequence.
Preferably, in step S3, the determining, by the text convolutional neural network, a cause of the failure includes:
s31, determiningDefining a plurality of one-dimensional convolution kernels, and then performing convolution calculation on input by using the convolution kernels, wherein the convolution kernels with different lengths can extract the mutual relation between a plurality of adjacent characters and words; assuming that input data of the model is a word vector matrix n, an input text consists of m words, weight in a neural network is w, B is offset, and obtaining characteristics AiThe calculation formula of (2) is as follows:
Ai＝F(w·ni:i+h-1+B)
in the formula: i ═ 1,2, …, m-h +1, h is the length of the convolution kernel, ". denotes the dot product operation, n represents the product of the valuesi:i+h-1Is the ith to the (i + h-1) th rows of the matrix n, and F is a non-linear function;
s32, replacing the representative features of all output channels by using time sequence maximum pooling, and splicing the representative features into a new feature vector;
s33, outputting the spliced vector as a label with the maximum probability through the full-connection layer, and updating the parameters by using a gradient descent algorithm through calculating a loss function of the prediction and the actual label;
and S34, when the power grid fault occurs, the knowledge graph is matched with a corresponding fault handling plan by sensing power grid alarm information and a text similarity technology, each operation event and operation event intention of the fault handling plan are presented, and the plan event is executed according to the corresponding intention to complete the fault handling.
Preferably, the text convolutional neural network automatically performs combination screening on the input features, and efficiently extracts important features through a convolutional layer and a pooling layer to obtain semantic information of different abstraction levels; according to the characteristic division, fault handling events are divided into a notification report class, a switch disconnecting link operation class, a regulating unit output class, a control device and section current class, a start-stop class, a control power grid voltage class and a control power grid frequency class.
Preferably, in step S4, the knowledge graph is automatically started when sensing the distribution grid fault, and drives the implementation of the whole disposal process by interacting with a dispatcher, and the human-computer interaction mode supports voice interaction or information pop-screen interaction; under the condition that the perceived fault is consistent with or similar to the fault in the plan or the case, guiding a knowledge graph to be carried out according to the process, guiding a dispatcher to carry out next treatment by the graph through perceiving power grid state calculation information and dispatcher decision conditions, wherein each treatment node is provided with a driving condition of a next node; the knowledge graph can support a dispatcher to inquire regulation regulations, notice items and contact ways of related personnel, a plurality of units are involved in the treatment process, a cooperative treatment tool is built, and information cooperative transmission, automatic clustering and automatic call making for the related personnel can be supported; and meanwhile, an automatic information recording function is established.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention provides a fault handling plan event extraction method based on the fusion of Albert-CRF and verb trigger words, and the intelligent extraction of fault handling knowledge is realized;
2. the invention provides an automatic execution method of a fault handling plan, which comprises the steps of establishing fault handling plan knowledge as a knowledge map, forming a handling event corresponding to a fault by using a knowledge map reasoning mechanism, establishing a plan event intention identification model by TextCNN, judging a plan event execution intention, and finishing automatic execution of the fault handling plan according to an identified plan event operation component and the plan event intention, namely finishing intelligent handling of distribution network faults.
Drawings
The invention is further described below with reference to the accompanying drawings.
Fig. 1 is a schematic diagram of the module connection of the present invention.
Detailed Description
As shown, in the first embodiment: the power distribution network fault handling knowledge map construction and intelligent aid decision making system comprises a fault handling data preprocessing module, a power distribution network fault handling knowledge extraction module, a fault handling knowledge map construction module and a fault handling decision reasoning module, wherein the fault handling data preprocessing module is connected with the power distribution network fault handling knowledge map extraction module, the power distribution network fault handling knowledge map construction module is connected with the fault handling knowledge map construction module, and the fault handling knowledge map construction module is connected with the fault handling decision reasoning module.
The power distribution network fault disposal knowledge graph construction and intelligent assistant decision method comprises the following steps:
the fault handling data preprocessing module cleans structured data and unstructured data based on regular expressions, wherein the structured data is various table information from a relational database and comprises the following steps: station, circuit, transformer, generating line, switch, circuit breaker, electric power company address book, new forms of energy station etc. table information, unstructured data contains: scheduling rules, fault handling plans, fault handling cases and other text information.
And secondly, extracting the fault disposal knowledge in the fault disposal plan by the power distribution network fault disposal knowledge extraction module based on the hybrid neural network to obtain fault disposal key points and key point relations. A distribution network fault handling plan BIO marking method and an event extraction method with verbs as trigger words are provided.
Firstly, classifying each event component, namely entity category, in the fault handling plan event text according to the natural grammar. The failure handling plan entity is divided into 6 categories of an execution subject, a verb trigger word, an execution object, an execution guest complement, an execution subject and an execution condition, 1 kind of labels of non-entities and 7 kinds of labels in total, and the specific representation mode is as follows: the initial part of the execution subject entity is represented as 'B-sub', and the rest part of the execution subject entity is represented as 'I-sub'; the initial part of the verb trigger entity is represented as 'B-act', and the rest of the verb trigger entity is represented as 'I-act'; the beginning part of the execution object entity is represented as 'B-obj', and the rest part of the execution object entity is represented as 'I-obj'; the initial part of the execution object is represented as 'B-obj _ m', and the rest part of the execution object is represented as 'I-obj _ m'; the beginning part of the executive shape language entity is represented as 'B-adv', and the rest part of the executive shape language entity is represented as 'I-adv'; the beginning part of the execution conditional entity is denoted as "B-con", the rest of the execution conditional entity is denoted as "I-con", and the non-entity component is denoted as "O".
And (3) adopting ALBERT-CRF to identify entity types in the fault handling plan, and combining 6 fault handling plan entities of a subject execution language, a verb trigger word, an object execution language, a guest supplement execution language, a shape execution language and an execution condition according to a grammatical relation to form a complete event, thus completing the extraction of the fault handling plan event. And converting the fault handling plan entity into a calculable word vector by using an ALBERT dynamic word vector model, realizing coding of the fault handling plan entity, sending the coded word vector into an ALBERT network parameter adjusting layer to predict the entity information label type of the character, and finally outputting a global optimal label sequence in a CRF layer by taking sentences as units.
The ALBERT is a light-weight BERT model, reduces the parameter dimension of a BERT hidden layer by decomposing word vectors, and can conceal the decomposed word vectors to the hidden layer according to requirements. The BERT model has huge parameter quantity of a self-attention layer and a feedforward neural network layer, and the ALBERT adopts a parameter sharing mechanism in the aspect, compresses original 12-layer parameters into 1-layer parameters, reduces the complexity of model calculation parameters, but has slightly influenced performance. In order to compensate for the performance loss, measures such as removing a model dropout layer, expanding a depth model and a Sentence Order Prediction (SOP) pre-training strategy are adopted to replace the Next Sentence Prediction (NSP).
Based on the advantages of the ALBERT pre-training model, the word vector model is used as a word vector model to carry out word vector coding on the marked fault cause diagnosis related information.
The function of the CRF layer is to predict the optimal label scores of all categories from the global perspective, and the calculation formula is as follows:
in the formula:
are the adjacent state transition matrix distribution values in the CRF,
is a fractional value of the LSTM model output vector. Attributing point values toAfter normalization, a probability formula for the model is obtained:
in the formula: YX denotes all sequence tags.
The CRF optimizes the label sequence by solving the logarithm of the probability, and finally obtains the sequence with the highest score as the final predicted output sequence through the maximum likelihood estimation, wherein the calculation formula is as follows:
in the formula:
represents the score of the entire predicted sequence.
Thirdly, the fault handling knowledge map construction module stores the extracted fault handling plan handling knowledge into a Neo4j database according to information categories such as plan titles, operation modes, control measures, fault reasons, control requirements, influences after faults and the like, modularizes the whole fault handling knowledge map, judges the operation intention of the combined plan event through a text convolutional neural network (TextCNN), and divides the operation intention into: notification reporting, switch disconnecting link operation, adjusting unit output, control equipment and section tide current, start-stop, control power grid voltage and control power grid frequency.
The TextCNN is divided into the following steps when fault cause diagnosis is performed:
(1) defining a plurality of one-dimensional convolution kernels, and performing convolution calculation on input by using the convolution kernels, wherein the convolution kernels with different lengths can extract the correlation among a plurality of adjacent characters (words). Input data of the hypothetical model isA word vector matrix n, wherein the input text consists of m characters, the weight in the neural network is w, B is offset, and the characteristic A is obtainediThe calculation formula of (2) is as follows:
Ai＝F(w·ni:i+h-1+B)
in the formula: i ═ 1,2, …, m-h +1, h is the length of the convolution kernel, ". denotes the dot product operation, n represents the product of the valuesi:i+h-1Is the ith to the (i + h-1) th rows of the matrix n, and F is a non-linear function;
(2) using time sequence maximum pooling to replace the representative features of all output channels, and then splicing the representative features into a new feature vector;
(3) outputting the spliced vector as a label with the maximum probability through the full-connection layer, and updating the parameters by using a gradient descent algorithm through calculating a loss function of the prediction and the actual label.
The text matrix characteristics are vector matrixes formed by word vectors identified by fault reasons, the TextCNN automatically performs combination screening on the input characteristics, and performs efficient extraction on important characteristics through a convolution layer and a pooling layer to obtain semantic information of different abstraction layers. According to the characteristic division, fault handling events are divided into a notification report class, a switch disconnecting link operation class, a regulating unit output class, a control device and section current class, a start-stop class, a control power grid voltage class and a control power grid frequency class.
When a power grid fault occurs, the knowledge graph is matched with a corresponding fault handling plan through sensing power grid alarm information and a text similarity technology, each operation event and operation event intention of the fault handling plan are presented, and the plan event is executed according to the corresponding intention to complete the fault handling.
And fourthly, the fault handling decision reasoning module takes the fault handling plan knowledge graph as a handling engine, and completes intelligent handling of the fault under the action of the reasoning machine by sensing the real-time running state of the power distribution network. The knowledge graph is automatically started when sensing the distribution power grid fault, the realization of the whole disposal process is driven through the interaction with a dispatcher, and the voice interaction or the information popup screen interaction is supported in a man-machine interaction mode. And under the condition that the perceived fault is consistent with or similar to the fault in the plan or the case, guiding a knowledge graph to be carried out according to the process, guiding a dispatcher to carry out next treatment by the graph through perceiving the power grid state calculation information and the decision condition of the dispatcher, and setting a driving condition of a next node by each treatment node. The knowledge graph can support a dispatcher to inquire regulation regulations, notice items and contact ways of related personnel, and in addition, a plurality of units are involved in the treatment process, and a cooperative treatment tool is built, so that the cooperative transmission of information, automatic clustering and automatic telephone dialing for the related personnel can be supported. Meanwhile, an automatic information recording function is built, and the recording speed of information and the tracing of the disposal information are improved.
The above embodiments are merely illustrative of the principles of the present invention and its effects, and do not limit the present invention. It will be apparent to those skilled in the art that modifications and improvements can be made to the above-described embodiments without departing from the spirit and scope of the invention. Accordingly, it is intended that all equivalent modifications or changes be made by those skilled in the art without departing from the spirit and technical spirit of the present invention, and be covered by the claims of the present invention.
Claims (10)
1. Distribution network fault handles knowledge map construction, intelligent aid decision-making system, its characterized in that: the fault treatment data preprocessing module is connected with the power distribution network fault treatment knowledge extraction module, the power distribution network fault treatment knowledge map construction module is connected with the fault treatment knowledge map construction module, and the fault treatment knowledge map construction module is connected with the fault treatment decision inference module.
2. The power distribution network fault handling knowledge graph construction and intelligent aid decision method based on the claim 1 is characterized by comprising the following steps:
s1, the fault handling data preprocessing module cleans the structured data and the unstructured data based on the regular expression;
s2, extracting fault disposal knowledge in a fault disposal plan by a power distribution network fault disposal knowledge extraction module based on a hybrid neural network to obtain fault disposal key points and key point relations, and extracting BIO marks and events of the power distribution network fault disposal plan by using verbs as trigger words;
s3, the fault handling knowledge map construction module stores the extracted fault handling plan handling knowledge into a Neo4j database according to information categories such as plan titles, operation modes, control measures, fault reasons, control requirements, after-fault influences and the like, modularizes the whole fault handling knowledge map, and judges the operation intention of the combined plan event through a text convolution neural network;
s4, the fault handling decision reasoning module takes a fault handling plan knowledge graph as a handling engine, and intelligent handling of faults is completed under the action of the reasoning engine by sensing the real-time running state of the power distribution network.
3. The power distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 2, characterized in that: in step S1, the structured data is table information from a relational database, and includes a plant station, a line, a transformer, a bus, a disconnecting link, a circuit breaker, an address book of a power company, and a new energy plant station, and the unstructured data is text information, and includes a scheduling rule, a fault handling plan, and a fault handling case.
4. The power distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 2, characterized in that: the step S2 includes the steps of:
s21, classifying each event component in the fault handling plan event text according to natural grammar, namely entity categories, and dividing the fault handling plan entity into 6 categories of execution subject, verb trigger, execution object, execution supplement, execution subject and execution condition, 1 non-entity label and 7 labels in total;
s22, adopting ALBERT-CRF to identify entity types in the fault handling plan, combining 6 fault handling plan entities of subject execution, verb trigger words, object execution, shape execution and condition execution to form a complete event according to grammatical relations, and finishing the extraction of the fault handling plan event;
s23, converting the fault handling plan entity into a calculable word vector by using an ALBERT dynamic word vector model, realizing coding of the fault handling plan entity, sending the coded word vector into entity information label types of predicted characters of an ALBERT network parameter adjusting layer, and finally outputting a global optimal label sequence by taking sentences as units on a CRF layer.
5. The distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 4, characterized in that: the initial part of the execution subject entity is represented as 'B-sub', and the rest part of the execution subject entity is represented as 'I-sub'; the initial part of the verb trigger entity is represented as 'B-act', and the rest of the verb trigger entity is represented as 'I-act'; the beginning part of the execution object entity is represented as 'B-obj', and the rest part of the execution object entity is represented as 'I-obj'; the initial part of the execution object is represented as 'B-obj _ m', and the rest part of the execution object is represented as 'I-obj _ m'; the beginning part of the executive shape language entity is represented as 'B-adv', and the rest part of the executive shape language entity is represented as 'I-adv'; the beginning part of the execution conditional entity is denoted as "B-con", the rest of the execution conditional entity is denoted as "I-con", and the non-entity component is denoted as "O".
6. The distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 4, characterized in that: the ALBERT is a light-weight BERT model, reduces the parameter dimension of a BERT hidden layer by decomposing word vectors, and can conceal the decomposed word vectors to the hidden layer according to requirements.
7. The distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 4, characterized in that: the function of the CRF layer is to predict the optimal label scores of all categories from the global perspective, and the calculation formula is as follows:
in the formula:
are the adjacent state transition matrix distribution values in the CRF,
is the fraction value of the LSTM model output vector, and after normalizing the fraction value, the probability formula of the model is obtained:
in the formula: YX represents all sequence tags;
the CRF optimizes the label sequence by solving the logarithm of the probability, and finally obtains the sequence with the highest score as the final predicted output sequence through the maximum likelihood estimation, wherein the calculation formula is as follows:
in the formula:
represents the score of the entire predicted sequence.
8. The power distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 2, characterized in that: in step S3, the determining, by the text convolutional neural network, a cause of the failure includes:
s31, defining a plurality of one-dimensional convolution kernels, then using the convolution kernels to perform convolution calculation on input, wherein the convolution kernels with different lengths can extract the interrelations between a plurality of adjacent characters and words; assuming that input data of the model is a word vector matrix n, an input text consists of m words, weight in a neural network is w, B is offset, and obtaining characteristics AiThe calculation formula of (2) is as follows:
Ai＝F(w·ni:i+h-1+B)
in the formula: i ═ 1,2, …, m-h +1, h is the length of the convolution kernel, ". denotes the dot product operation, n represents the product of the valuesi:i+h-1Is the ith to the (i + h-1) th rows of the matrix n, and F is a non-linear function;
s32, replacing the representative features of all output channels by using time sequence maximum pooling, and splicing the representative features into a new feature vector;
s33, outputting the spliced vector as a label with the maximum probability through the full-connection layer, and updating the parameters by using a gradient descent algorithm through calculating a loss function of the prediction and the actual label;
and S34, when the power grid fault occurs, the knowledge graph is matched with a corresponding fault handling plan by sensing power grid alarm information and a text similarity technology, each operation event and operation event intention of the fault handling plan are presented, and the plan event is executed according to the corresponding intention to complete the fault handling.
9. The distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 8, characterized in that: the text convolutional neural network automatically performs combined screening on input features, and performs high-efficiency extraction on important features through a convolutional layer and a pooling layer to obtain semantic information of different abstract levels; according to the characteristic division, fault handling events are divided into a notification report class, a switch disconnecting link operation class, a regulating unit output class, a control device and section current class, a start-stop class, a control power grid voltage class and a control power grid frequency class.
10. The power distribution network fault handling knowledge graph construction and intelligent aid decision method according to claim 2, characterized in that: in the step S4, the knowledge graph is automatically started when sensing the distribution grid fault, and drives the realization of the whole disposal process by interacting with the dispatcher, and the human-computer interaction mode supports voice interaction or information pop-screen interaction; under the condition that the perceived fault is consistent with or similar to the fault in the plan or the case, guiding a knowledge graph to be carried out according to the process, guiding a dispatcher to carry out next treatment by the graph through perceiving power grid state calculation information and dispatcher decision conditions, wherein each treatment node is provided with a driving condition of a next node; the knowledge graph can support a dispatcher to inquire regulation regulations, notice items and contact ways of related personnel, a plurality of units are involved in the treatment process, a cooperative treatment tool is built, and information cooperative transmission, automatic clustering and automatic call making for the related personnel can be supported; and meanwhile, an automatic information recording function is established.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111465857.7A CN114328950A (en) | 2021-12-03 | 2021-12-03 | Power distribution network fault disposal knowledge map construction and intelligent aid decision making system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111465857.7A CN114328950A (en) | 2021-12-03 | 2021-12-03 | Power distribution network fault disposal knowledge map construction and intelligent aid decision making system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114328950A true CN114328950A (en) | 2022-04-12 |
Family
ID=81047993
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111465857.7A Pending CN114328950A (en) | 2021-12-03 | 2021-12-03 | Power distribution network fault disposal knowledge map construction and intelligent aid decision making system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114328950A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115099338A (en) * | 2022-06-24 | 2022-09-23 | 国网浙江省电力有限公司电力科学研究院 | Power grid master equipment-oriented multi-source heterogeneous quality information fusion processing method and system |
| CN115292518A (en) * | 2022-08-18 | 2022-11-04 | 国家电网有限公司 | Power distribution network fault processing method and system based on knowledge type information extraction |
| CN116091045A (en) * | 2023-02-28 | 2023-05-09 | 武汉烽火技术服务有限公司 | Knowledge-graph-based communication network operation and maintenance method and operation and maintenance device |
| CN117171332A (en) * | 2023-11-02 | 2023-12-05 | 江西拓世智能科技股份有限公司 | Intelligent question-answering method and system based on AI |
| CN117196354A (en) * | 2023-11-08 | 2023-12-08 | 国网浙江省电力有限公司 | Intelligent decision method for multi-mode perception and domain map model |
| CN117273375A (en) * | 2023-10-19 | 2023-12-22 | 国网安徽省电力有限公司铜陵供电公司 | Distribution network fault handling decision supervision and lifting system based on knowledge graph |
-
2021
- 2021-12-03 CN CN202111465857.7A patent/CN114328950A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115099338A (en) * | 2022-06-24 | 2022-09-23 | 国网浙江省电力有限公司电力科学研究院 | Power grid master equipment-oriented multi-source heterogeneous quality information fusion processing method and system |
| CN115292518A (en) * | 2022-08-18 | 2022-11-04 | 国家电网有限公司 | Power distribution network fault processing method and system based on knowledge type information extraction |
| CN116091045A (en) * | 2023-02-28 | 2023-05-09 | 武汉烽火技术服务有限公司 | Knowledge-graph-based communication network operation and maintenance method and operation and maintenance device |
| CN117273375A (en) * | 2023-10-19 | 2023-12-22 | 国网安徽省电力有限公司铜陵供电公司 | Distribution network fault handling decision supervision and lifting system based on knowledge graph |
| CN117273375B (en) * | 2023-10-19 | 2024-04-02 | 国网安徽省电力有限公司铜陵供电公司 | Distribution network fault handling decision supervision and lifting system based on knowledge graph |
| CN117171332A (en) * | 2023-11-02 | 2023-12-05 | 江西拓世智能科技股份有限公司 | Intelligent question-answering method and system based on AI |
| CN117196354A (en) * | 2023-11-08 | 2023-12-08 | 国网浙江省电力有限公司 | Intelligent decision method for multi-mode perception and domain map model |
| CN117196354B (en) * | 2023-11-08 | 2024-01-30 | 国网浙江省电力有限公司 | Intelligent decision method for multi-mode perception and domain map model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114328950A (en) | Power distribution network fault disposal knowledge map construction and intelligent aid decision making system and method | |
| CN105591882B (en) | A kind of intelligence machine person to person mixes the method and system of customer service | |
| CN112419096B (en) | Automatic circulation method of user electricity appeal work order based on NLP information extraction and few-sample self-learning | |
| CN111475655B (en) | Power distribution network knowledge graph-based power scheduling text entity linking method | |
| CN111581350A (en) | Multi-task learning, reading and understanding method based on pre-training language model | |
| CN112632288A (en) | Power dispatching system and method based on knowledge graph | |
| CN110888973B (en) | Method for automatically structuring and carding monitoring information table | |
| CN116010581A (en) | Knowledge graph question-answering method and system based on power grid hidden trouble shooting scene | |
| CN114661905A (en) | Power grid fault diagnosis method based on BERT | |
| CN111191452A (en) | Railway text named entity recognition method and device | |
| CN115619117A (en) | Power grid intelligent scheduling method based on duty system | |
| CN112232078A (en) | Scheduling operation ticket auditing method based on bidirectional GRU and attention mechanism | |
| CN112036179A (en) | Electric power plan information extraction method based on text classification and semantic framework | |
| CN118260689A (en) | Log anomaly detection method based on high-efficiency fine adjustment of self-adaptive low-rank parameters | |
| CN118503494A (en) | Multi-source heterogeneous knowledge base-oriented large model question-answering method, device and equipment | |
| CN110782221A (en) | Intelligent interview evaluation system and method | |
| CN110175332A (en) | A kind of intelligence based on artificial neural network is set a question method and system | |
| CN117764774A (en) | Knowledge graph-based power grid dispatching risk analysis method, system and storage medium | |
| CN112287084A (en) | Question-answering method and system based on ensemble learning | |
| CN115438665A (en) | Power supply service work order early warning identification method, device, equipment and medium | |
| CN113487194A (en) | Electric power system dispatcher grade evaluation system based on text classification | |
| Hundal et al. | Product design development by integrating QFD approach with heuristics-AHP, ANN and fuzzy logics-a case study in miniature circuit breaker | |
| CN113434667B (en) | Text classification method based on distribution network automation terminal text classification model | |
| Xiang et al. | Design of Intelligent Dispatching System Based on Human Voice Adaptive Speech Recognition | |
| CN116561287B (en) | Intelligent customer service implementation method and system based on deep semantic learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |